On 26/02/2012 00:43, Mark Thomas wrote: > On 25/02/2012 13:32, ma...@apache.org wrote: >> Mark Thomas <ma...@apache.org> wrote: > >>>> 5. Threading. Fix the TODOs associated around multiple threads >>>> trying >>> to >>>> send messages at the same time. >> >> Still to do. >> >>>> 6. Profiling. Take a look with YourKit and Autobahn's >>>> performance >>> tests. >>>> If there are obvious bottlenecks, fix them. >> >> Mostly fixed. Performance for lots of small messages is still >> terrible but I haven't added any buffering yet and that should help a >> lot. There are a few obvious bottle necks I want to fix before i add >> the buffering. > > This turned out to be a Windows issue. Running the same tests on Linux > is significantly faster and on a par with the performance figures > published by Autobahn for various servers they tested. > > I'm not sure if it is Java, Python or both on Windows but something, > somewhere really doesn't like lots of small packets. Adding the > buffering made zero difference so I don't see that there is anything I > can do in the WebSocket code to help with this.
It is python on Windows that doesn't like small packets. Running the test suite on Linux and Tomcat on Windows gives similar performance to running everything on Windows. I'm now positive there is nothing we can do in Tomcat to fix this. Mark > Also on the down side, the WebSockets does not play nicely with the > APR/native connector yet. I need to dig into that and test the NIO > connector. > >>>> 7. Add some documentation but mainly rely on the examples and >>>> the >>> Javadoc. >> >> Still to do. > > Mark > > > --------------------------------------------------------------------- > To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org > For additional commands, e-mail: dev-h...@tomcat.apache.org > --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org For additional commands, e-mail: dev-h...@tomcat.apache.org
